projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
udp: remove headers from UDP packets before queueing
[deliverable/linux.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
eb4cb008 134#include <linux/sock_diag.h>
1da177e4
LT		 135
136#include <linux/filter.h>
538950a1 137#include <net/sock_reuseport.h>
1da177e4 138
3847ce32
SM		 139#include <trace/events/sock.h>
140
1da177e4
LT		 141#ifdef CONFIG_INET
142#include <net/tcp.h>
143#endif
144
076bb0c8 145#include <net/busy_poll.h>
06021292 146
36b77a52 147static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 148static LIST_HEAD(proto_list);
149
a3b299da
EB		 150/**
151 * sk_ns_capable - General socket capability test
152 * @sk: Socket to use a capability on or through
153 * @user_ns: The user namespace of the capability to use
154 * @cap: The capability to use
155 *
156 * Test to see if the opener of the socket had when the socket was
157 * created and the current process has the capability @cap in the user
158 * namespace @user_ns.
159 */
160bool sk_ns_capable(const struct sock *sk,
161 struct user_namespace *user_ns, int cap)
162{
163 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
164 ns_capable(user_ns, cap);
165}
166EXPORT_SYMBOL(sk_ns_capable);
167
168/**
169 * sk_capable - Socket global capability test
170 * @sk: Socket to use a capability on or through
e793c0f7 171 * @cap: The global capability to use
a3b299da
EB		 172 *
173 * Test to see if the opener of the socket had when the socket was
174 * created and the current process has the capability @cap in all user
175 * namespaces.
176 */
177bool sk_capable(const struct sock *sk, int cap)
178{
179 return sk_ns_capable(sk, &init_user_ns, cap);
180}
181EXPORT_SYMBOL(sk_capable);
182
183/**
184 * sk_net_capable - Network namespace socket capability test
185 * @sk: Socket to use a capability on or through
186 * @cap: The capability to use
187 *
e793c0f7 188 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 189 * and the current process has the capability @cap over the network namespace
190 * the socket is a member of.
191 */
192bool sk_net_capable(const struct sock *sk, int cap)
193{
194 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
195}
196EXPORT_SYMBOL(sk_net_capable);
197
da21f24d
IM		 198/*
199 * Each address family might have different locking rules, so we have
200 * one slock key per address family:
201 */
a5b5bb9a
IM		 202static struct lock_class_key af_family_keys[AF_MAX];
203static struct lock_class_key af_family_slock_keys[AF_MAX];
204
a5b5bb9a
IM		 205/*
206 * Make lock validator output more readable. (we pre-construct these
207 * strings build-time, so that runtime initialization of socket
208 * locks is fast):
209 */
36cbd3dc 210static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 211 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
212 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
213 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
214 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
215 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
216 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
217 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 218 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 219 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 220 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 221 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 222 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 223 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 224 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 225};
36cbd3dc 226static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 227 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
228 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
229 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
230 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
231 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
232 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
233 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 234 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 235 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 236 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 237 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 238 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 239 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 240 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 241};
36cbd3dc 242static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 243 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
244 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
245 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
246 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
247 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
248 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
249 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 250 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 251 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 252 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 253 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 254 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 255 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 256 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 257};
da21f24d
IM		 258
259/*
260 * sk_callback_lock locking rules are per-address-family,
261 * so split the lock classes by using a per-AF key:
262 */
263static struct lock_class_key af_callback_keys[AF_MAX];
264
1da177e4
LT		 265/* Take into consideration the size of the struct sk_buff overhead in the
266 * determination of these values, since that is non-constant across
267 * platforms. This makes socket queueing behavior and performance
268 * not depend upon such differences.
269 */
270#define _SK_MEM_PACKETS 256
87fb4b7b 271#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 272#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
273#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
274
275/* Run time adjustable parameters. */
ab32ea5d 276__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 277EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 278__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 279EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 280__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
281__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 282
25985edc 283/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 284int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 285EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 286
b245be1f
WB		 287int sysctl_tstamp_allow_data __read_mostly = 1;
288
c93bdd0e
MG		 289struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
290EXPORT_SYMBOL_GPL(memalloc_socks);
291
7cb02404
MG		 292/**
293 * sk_set_memalloc - sets %SOCK_MEMALLOC
294 * @sk: socket to set it on
295 *
296 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
297 * It's the responsibility of the admin to adjust min_free_kbytes
298 * to meet the requirements
299 */
300void sk_set_memalloc(struct sock *sk)
301{
302 sock_set_flag(sk, SOCK_MEMALLOC);
303 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 304 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 305}
306EXPORT_SYMBOL_GPL(sk_set_memalloc);
307
308void sk_clear_memalloc(struct sock *sk)
309{
310 sock_reset_flag(sk, SOCK_MEMALLOC);
311 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 312 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 313
314 /*
315 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 316 * progress of swapping. SOCK_MEMALLOC may be cleared while
317 * it has rmem allocations due to the last swapfile being deactivated
318 * but there is a risk that the socket is unusable due to exceeding
319 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 320 */
5d753610 321 sk_mem_reclaim(sk);
7cb02404
MG		 322}
323EXPORT_SYMBOL_GPL(sk_clear_memalloc);
324
b4b9e355
MG		 325int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
326{
327 int ret;
328 unsigned long pflags = current->flags;
329
330 /* these should have been dropped before queueing */
331 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
332
333 current->flags |= PF_MEMALLOC;
334 ret = sk->sk_backlog_rcv(sk, skb);
335 tsk_restore_flags(current, pflags, PF_MEMALLOC);
336
337 return ret;
338}
339EXPORT_SYMBOL(__sk_backlog_rcv);
340
1da177e4
LT		 341static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
342{
343 struct timeval tv;
344
345 if (optlen < sizeof(tv))
346 return -EINVAL;
347 if (copy_from_user(&tv, optval, sizeof(tv)))
348 return -EFAULT;
ba78073e
VA		 349 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
350 return -EDOM;
1da177e4 351
ba78073e 352 if (tv.tv_sec < 0) {
6f11df83
AM		 353 static int warned __read_mostly;
354
ba78073e 355 *timeo_p = 0;
50aab54f 356 if (warned < 10 && net_ratelimit()) {
ba78073e 357 warned++;
e005d193
JP		 358 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
359 __func__, current->comm, task_pid_nr(current));
50aab54f 360 }
ba78073e
VA		 361 return 0;
362 }
1da177e4
LT		 363 *timeo_p = MAX_SCHEDULE_TIMEOUT;
364 if (tv.tv_sec == 0 && tv.tv_usec == 0)
365 return 0;
366 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
367 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
368 return 0;
369}
370
371static void sock_warn_obsolete_bsdism(const char *name)
372{
373 static int warned;
374 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 375 if (strcmp(warncomm, current->comm) && warned < 5) {
376 strcpy(warncomm, current->comm);
e005d193
JP		 377 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
378 warncomm, name);
1da177e4
LT		 379 warned++;
380 }
381}
382
080a270f
HFS		 383static bool sock_needs_netstamp(const struct sock *sk)
384{
385 switch (sk->sk_family) {
386 case AF_UNSPEC:
387 case AF_UNIX:
388 return false;
389 default:
390 return true;
391 }
392}
393
08e29af3 394static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 395{
08e29af3
ED		 396 if (sk->sk_flags & flags) {
397 sk->sk_flags &= ~flags;
080a270f
HFS		 398 if (sock_needs_netstamp(sk) &&
399 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 400 net_disable_timestamp();
1da177e4
LT		 401 }
402}
403
404
e6afc8ac 405int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
f0088a50 406{
3b885787
NH		 407 unsigned long flags;
408 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 409
0fd7bac6 410 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 411 atomic_inc(&sk->sk_drops);
3847ce32 412 trace_sock_rcvqueue_full(sk, skb);
766e9037 413 return -ENOMEM;
f0088a50
DV		 414 }
415
c76562b6 416 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 417 atomic_inc(&sk->sk_drops);
418 return -ENOBUFS;
3ab224be
HA		 419 }
420
f0088a50
DV		 421 skb->dev = NULL;
422 skb_set_owner_r(skb, sk);
49ad9599 423
7fee226a
ED		 424 /* we escape from rcu protected region, make sure we dont leak
425 * a norefcounted dst
426 */
427 skb_dst_force(skb);
428
3b885787 429 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 430 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 431 __skb_queue_tail(list, skb);
432 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 433
434 if (!sock_flag(sk, SOCK_DEAD))
676d2369 435 sk->sk_data_ready(sk);
766e9037 436 return 0;
f0088a50 437}
e6afc8ac 438EXPORT_SYMBOL(__sock_queue_rcv_skb);
439
440int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
441{
442 int err;
443
444 err = sk_filter(sk, skb);
445 if (err)
446 return err;
447
448 return __sock_queue_rcv_skb(sk, skb);
449}
f0088a50
DV		 450EXPORT_SYMBOL(sock_queue_rcv_skb);
451
58a5a7b9 452int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 453{
454 int rc = NET_RX_SUCCESS;
455
fda9ef5d 456 if (sk_filter(sk, skb))
f0088a50
DV		 457 goto discard_and_relse;
458
459 skb->dev = NULL;
460
274f482d 461 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 462 atomic_inc(&sk->sk_drops);
463 goto discard_and_relse;
464 }
58a5a7b9
ACM		 465 if (nested)
466 bh_lock_sock_nested(sk);
467 else
468 bh_lock_sock(sk);
a5b5bb9a
IM		 469 if (!sock_owned_by_user(sk)) {
470 /*
471 * trylock + unlock semantics:
472 */
473 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
474
c57943a1 475 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 476
477 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 478 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 479 bh_unlock_sock(sk);
480 atomic_inc(&sk->sk_drops);
481 goto discard_and_relse;
482 }
483
f0088a50
DV		 484 bh_unlock_sock(sk);
485out:
486 sock_put(sk);
487 return rc;
488discard_and_relse:
489 kfree_skb(skb);
490 goto out;
491}
492EXPORT_SYMBOL(sk_receive_skb);
493
494struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
495{
b6c6712a 496 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 497
498 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 499 sk_tx_queue_clear(sk);
a9b3cd7f 500 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 501 dst_release(dst);
502 return NULL;
503 }
504
505 return dst;
506}
507EXPORT_SYMBOL(__sk_dst_check);
508
509struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
510{
511 struct dst_entry *dst = sk_dst_get(sk);
512
513 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
514 sk_dst_reset(sk);
515 dst_release(dst);
516 return NULL;
517 }
518
519 return dst;
520}
521EXPORT_SYMBOL(sk_dst_check);
522
c91f6df2
BH		 523static int sock_setbindtodevice(struct sock *sk, char __user *optval,
524 int optlen)
4878809f
DM		 525{
526 int ret = -ENOPROTOOPT;
527#ifdef CONFIG_NETDEVICES
3b1e0a65 528 struct net *net = sock_net(sk);
4878809f
DM		 529 char devname[IFNAMSIZ];
530 int index;
531
532 /* Sorry... */
533 ret = -EPERM;
5e1fccc0 534 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 535 goto out;
536
537 ret = -EINVAL;
538 if (optlen < 0)
539 goto out;
540
541 /* Bind this socket to a particular device like "eth0",
542 * as specified in the passed interface name. If the
543 * name is "" or the option length is zero the socket
544 * is not bound.
545 */
546 if (optlen > IFNAMSIZ - 1)
547 optlen = IFNAMSIZ - 1;
548 memset(devname, 0, sizeof(devname));
549
550 ret = -EFAULT;
551 if (copy_from_user(devname, optval, optlen))
552 goto out;
553
000ba2e4
DM		 554 index = 0;
555 if (devname[0] != '\0') {
bf8e56bf 556 struct net_device *dev;
4878809f 557
bf8e56bf
ED		 558 rcu_read_lock();
559 dev = dev_get_by_name_rcu(net, devname);
560 if (dev)
561 index = dev->ifindex;
562 rcu_read_unlock();
4878809f
DM		 563 ret = -ENODEV;
564 if (!dev)
565 goto out;
4878809f
DM		 566 }
567
568 lock_sock(sk);
569 sk->sk_bound_dev_if = index;
570 sk_dst_reset(sk);
571 release_sock(sk);
572
573 ret = 0;
574
575out:
576#endif
577
578 return ret;
579}
580
c91f6df2
BH		 581static int sock_getbindtodevice(struct sock *sk, char __user *optval,
582 int __user *optlen, int len)
583{
584 int ret = -ENOPROTOOPT;
585#ifdef CONFIG_NETDEVICES
586 struct net *net = sock_net(sk);
c91f6df2 587 char devname[IFNAMSIZ];
c91f6df2
BH		 588
589 if (sk->sk_bound_dev_if == 0) {
590 len = 0;
591 goto zero;
592 }
593
594 ret = -EINVAL;
595 if (len < IFNAMSIZ)
596 goto out;
597
5dbe7c17
NS		 598 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
599 if (ret)
c91f6df2 600 goto out;
c91f6df2
BH		 601
602 len = strlen(devname) + 1;
603
604 ret = -EFAULT;
605 if (copy_to_user(optval, devname, len))
606 goto out;
607
608zero:
609 ret = -EFAULT;
610 if (put_user(len, optlen))
611 goto out;
612
613 ret = 0;
614
615out:
616#endif
617
618 return ret;
619}
620
c0ef877b
PE		 621static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
622{
623 if (valbool)
624 sock_set_flag(sk, bit);
625 else
626 sock_reset_flag(sk, bit);
627}
628
f60e5990 629bool sk_mc_loop(struct sock *sk)
630{
631 if (dev_recursion_level())
632 return false;
633 if (!sk)
634 return true;
635 switch (sk->sk_family) {
636 case AF_INET:
637 return inet_sk(sk)->mc_loop;
638#if IS_ENABLED(CONFIG_IPV6)
639 case AF_INET6:
640 return inet6_sk(sk)->mc_loop;
641#endif
642 }
643 WARN_ON(1);
644 return true;
645}
646EXPORT_SYMBOL(sk_mc_loop);
647
1da177e4
LT		 648/*
649 * This is meant for all protocols to use and covers goings on
650 * at the socket level. Everything here is generic.
651 */
652
653int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 654 char __user *optval, unsigned int optlen)
1da177e4 655{
2a91525c 656 struct sock *sk = sock->sk;
1da177e4
LT		 657 int val;
658 int valbool;
659 struct linger ling;
660 int ret = 0;
4ec93edb 661
1da177e4
LT		 662 /*
663 * Options without arguments
664 */
665
4878809f 666 if (optname == SO_BINDTODEVICE)
c91f6df2 667 return sock_setbindtodevice(sk, optval, optlen);
4878809f 668
e71a4783
SH		 669 if (optlen < sizeof(int))
670 return -EINVAL;
4ec93edb 671
1da177e4
LT		 672 if (get_user(val, (int __user *)optval))
673 return -EFAULT;
4ec93edb 674
2a91525c 675 valbool = val ? 1 : 0;
1da177e4
LT		 676
677 lock_sock(sk);
678
2a91525c 679 switch (optname) {
e71a4783 680 case SO_DEBUG:
2a91525c 681 if (val && !capable(CAP_NET_ADMIN))
e71a4783 682 ret = -EACCES;
2a91525c 683 else
c0ef877b 684 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 685 break;
686 case SO_REUSEADDR:
4a17fd52 687 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 688 break;
055dc21a
TH		 689 case SO_REUSEPORT:
690 sk->sk_reuseport = valbool;
691 break;
e71a4783 692 case SO_TYPE:
49c794e9 693 case SO_PROTOCOL:
0d6038ee 694 case SO_DOMAIN:
e71a4783
SH		 695 case SO_ERROR:
696 ret = -ENOPROTOOPT;
697 break;
698 case SO_DONTROUTE:
c0ef877b 699 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 700 break;
701 case SO_BROADCAST:
702 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
703 break;
704 case SO_SNDBUF:
705 /* Don't error on this BSD doesn't and if you think
82981930
ED		 706 * about it this is right. Otherwise apps have to
707 * play 'guess the biggest size' games. RCVBUF/SNDBUF
708 * are treated in BSD as hints
709 */
710 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 711set_sndbuf:
e71a4783 712 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 713 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
714 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 715 sk->sk_write_space(sk);
716 break;
1da177e4 717
e71a4783
SH		 718 case SO_SNDBUFFORCE:
719 if (!capable(CAP_NET_ADMIN)) {
720 ret = -EPERM;
721 break;
722 }
723 goto set_sndbuf;
b0573dea 724
e71a4783
SH		 725 case SO_RCVBUF:
726 /* Don't error on this BSD doesn't and if you think
82981930
ED		 727 * about it this is right. Otherwise apps have to
728 * play 'guess the biggest size' games. RCVBUF/SNDBUF
729 * are treated in BSD as hints
730 */
731 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 732set_rcvbuf:
e71a4783
SH		 733 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
734 /*
735 * We double it on the way in to account for
736 * "struct sk_buff" etc. overhead. Applications
737 * assume that the SO_RCVBUF setting they make will
738 * allow that much actual data to be received on that
739 * socket.
740 *
741 * Applications are unaware that "struct sk_buff" and
742 * other overheads allocate from the receive buffer
743 * during socket buffer allocation.
744 *
745 * And after considering the possible alternatives,
746 * returning the value we actually used in getsockopt
747 * is the most desirable behavior.
748 */
82981930 749 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 750 break;
751
752 case SO_RCVBUFFORCE:
753 if (!capable(CAP_NET_ADMIN)) {
754 ret = -EPERM;
1da177e4 755 break;
e71a4783
SH		 756 }
757 goto set_rcvbuf;
1da177e4 758
e71a4783 759 case SO_KEEPALIVE:
1da177e4 760#ifdef CONFIG_INET
3e10986d
ED		 761 if (sk->sk_protocol == IPPROTO_TCP &&
762 sk->sk_type == SOCK_STREAM)
e71a4783 763 tcp_set_keepalive(sk, valbool);
1da177e4 764#endif
e71a4783
SH		 765 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
766 break;
767
768 case SO_OOBINLINE:
769 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
770 break;
771
772 case SO_NO_CHECK:
28448b80 773 sk->sk_no_check_tx = valbool;
e71a4783
SH		 774 break;
775
776 case SO_PRIORITY:
5e1fccc0
EB		 777 if ((val >= 0 && val <= 6) ||
778 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 779 sk->sk_priority = val;
780 else
781 ret = -EPERM;
782 break;
783
784 case SO_LINGER:
785 if (optlen < sizeof(ling)) {
786 ret = -EINVAL; /* 1003.1g */
1da177e4 787 break;
e71a4783 788 }
2a91525c 789 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 790 ret = -EFAULT;
1da177e4 791 break;
e71a4783
SH		 792 }
793 if (!ling.l_onoff)
794 sock_reset_flag(sk, SOCK_LINGER);
795 else {
1da177e4 796#if (BITS_PER_LONG == 32)
e71a4783
SH		 797 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
798 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 799 else
e71a4783
SH		 800#endif
801 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
802 sock_set_flag(sk, SOCK_LINGER);
803 }
804 break;
805
806 case SO_BSDCOMPAT:
807 sock_warn_obsolete_bsdism("setsockopt");
808 break;
809
810 case SO_PASSCRED:
811 if (valbool)
812 set_bit(SOCK_PASSCRED, &sock->flags);
813 else
814 clear_bit(SOCK_PASSCRED, &sock->flags);
815 break;
816
817 case SO_TIMESTAMP:
92f37fd2 818 case SO_TIMESTAMPNS:
e71a4783 819 if (valbool) {
92f37fd2
ED		 820 if (optname == SO_TIMESTAMP)
821 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
822 else
823 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 824 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 825 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 826 } else {
e71a4783 827 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 828 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
829 }
e71a4783
SH		 830 break;
831
20d49473
PO		 832 case SO_TIMESTAMPING:
833 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 834 ret = -EINVAL;
20d49473
PO		 835 break;
836 }
b245be1f 837
09c2d251 838 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765 839 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 840 if (sk->sk_protocol == IPPROTO_TCP &&
841 sk->sk_type == SOCK_STREAM) {
6db8b963
SHY		 842 if ((1 << sk->sk_state) &
843 (TCPF_CLOSE | TCPF_LISTEN)) {
4ed2d765
WB		 844 ret = -EINVAL;
845 break;
846 }
847 sk->sk_tskey = tcp_sk(sk)->snd_una;
848 } else {
849 sk->sk_tskey = 0;
850 }
851 }
b9f40e21 852 sk->sk_tsflags = val;
20d49473
PO		 853 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
854 sock_enable_timestamp(sk,
855 SOCK_TIMESTAMPING_RX_SOFTWARE);
856 else
857 sock_disable_timestamp(sk,
08e29af3 858 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 859 break;
860
e71a4783
SH		 861 case SO_RCVLOWAT:
862 if (val < 0)
863 val = INT_MAX;
864 sk->sk_rcvlowat = val ? : 1;
865 break;
866
867 case SO_RCVTIMEO:
868 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
869 break;
870
871 case SO_SNDTIMEO:
872 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
873 break;
1da177e4 874
e71a4783
SH		 875 case SO_ATTACH_FILTER:
876 ret = -EINVAL;
877 if (optlen == sizeof(struct sock_fprog)) {
878 struct sock_fprog fprog;
1da177e4 879
e71a4783
SH		 880 ret = -EFAULT;
881 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 882 break;
e71a4783
SH		 883
884 ret = sk_attach_filter(&fprog, sk);
885 }
886 break;
887
89aa0758
AS		 888 case SO_ATTACH_BPF:
889 ret = -EINVAL;
890 if (optlen == sizeof(u32)) {
891 u32 ufd;
892
893 ret = -EFAULT;
894 if (copy_from_user(&ufd, optval, sizeof(ufd)))
895 break;
896
897 ret = sk_attach_bpf(ufd, sk);
898 }
899 break;
900
538950a1
CG		 901 case SO_ATTACH_REUSEPORT_CBPF:
902 ret = -EINVAL;
903 if (optlen == sizeof(struct sock_fprog)) {
904 struct sock_fprog fprog;
905
906 ret = -EFAULT;
907 if (copy_from_user(&fprog, optval, sizeof(fprog)))
908 break;
909
910 ret = sk_reuseport_attach_filter(&fprog, sk);
911 }
912 break;
913
914 case SO_ATTACH_REUSEPORT_EBPF:
915 ret = -EINVAL;
916 if (optlen == sizeof(u32)) {
917 u32 ufd;
918
919 ret = -EFAULT;
920 if (copy_from_user(&ufd, optval, sizeof(ufd)))
921 break;
922
923 ret = sk_reuseport_attach_bpf(ufd, sk);
924 }
925 break;
926
e71a4783 927 case SO_DETACH_FILTER:
55b33325 928 ret = sk_detach_filter(sk);
e71a4783 929 break;
1da177e4 930
d59577b6
VB		 931 case SO_LOCK_FILTER:
932 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
933 ret = -EPERM;
934 else
935 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
936 break;
937
e71a4783
SH		 938 case SO_PASSSEC:
939 if (valbool)
940 set_bit(SOCK_PASSSEC, &sock->flags);
941 else
942 clear_bit(SOCK_PASSSEC, &sock->flags);
943 break;
4a19ec58 944 case SO_MARK:
5e1fccc0 945 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 946 ret = -EPERM;
2a91525c 947 else
4a19ec58 948 sk->sk_mark = val;
4a19ec58 949 break;
877ce7c1 950
3b885787 951 case SO_RXQ_OVFL:
8083f0fc 952 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 953 break;
6e3e939f
JB		 954
955 case SO_WIFI_STATUS:
956 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
957 break;
958
ef64a54f
PE		 959 case SO_PEEK_OFF:
960 if (sock->ops->set_peek_off)
12663bfc 961 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 962 else
963 ret = -EOPNOTSUPP;
964 break;
3bdc0eba
BG		 965
966 case SO_NOFCS:
967 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
968 break;
969
7d4c04fc
KJ		 970 case SO_SELECT_ERR_QUEUE:
971 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
972 break;
973
e0d1095a 974#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 975 case SO_BUSY_POLL:
dafcc438
ET		 976 /* allow unprivileged users to decrease the value */
977 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
978 ret = -EPERM;
979 else {
980 if (val < 0)
981 ret = -EINVAL;
982 else
983 sk->sk_ll_usec = val;
984 }
985 break;
986#endif
62748f32
ED		 987
988 case SO_MAX_PACING_RATE:
989 sk->sk_max_pacing_rate = val;
990 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
991 sk->sk_max_pacing_rate);
992 break;
993
70da268b
ED		 994 case SO_INCOMING_CPU:
995 sk->sk_incoming_cpu = val;
996 break;
997
a87cb3e4
TH		 998 case SO_CNX_ADVICE:
999 if (val == 1)
1000 dst_negative_advice(sk);
1001 break;
e71a4783
SH		 1002 default:
1003 ret = -ENOPROTOOPT;
1004 break;
4ec93edb 1005 }
1da177e4
LT		 1006 release_sock(sk);
1007 return ret;
1008}
2a91525c 1009EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1010
1011
8f09898b 1012static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1013 struct ucred *ucred)
3f551f94
EB		 1014{
1015 ucred->pid = pid_vnr(pid);
1016 ucred->uid = ucred->gid = -1;
1017 if (cred) {
1018 struct user_namespace *current_ns = current_user_ns();
1019
b2e4f544
EB		 1020 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1021 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1022 }
1023}
1024
1da177e4
LT		 1025int sock_getsockopt(struct socket *sock, int level, int optname,
1026 char __user *optval, int __user *optlen)
1027{
1028 struct sock *sk = sock->sk;
4ec93edb 1029
e71a4783 1030 union {
4ec93edb
YH		 1031 int val;
1032 struct linger ling;
1da177e4
LT		 1033 struct timeval tm;
1034 } v;
4ec93edb 1035
4d0392be 1036 int lv = sizeof(int);
1da177e4 1037 int len;
4ec93edb 1038
e71a4783 1039 if (get_user(len, optlen))
4ec93edb 1040 return -EFAULT;
e71a4783 1041 if (len < 0)
1da177e4 1042 return -EINVAL;
4ec93edb 1043
50fee1de 1044 memset(&v, 0, sizeof(v));
df0bca04 1045
2a91525c 1046 switch (optname) {
e71a4783
SH		 1047 case SO_DEBUG:
1048 v.val = sock_flag(sk, SOCK_DBG);
1049 break;
1050
1051 case SO_DONTROUTE:
1052 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1053 break;
1054
1055 case SO_BROADCAST:
1b23a5df 1056 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1057 break;
1058
1059 case SO_SNDBUF:
1060 v.val = sk->sk_sndbuf;
1061 break;
1062
1063 case SO_RCVBUF:
1064 v.val = sk->sk_rcvbuf;
1065 break;
1066
1067 case SO_REUSEADDR:
1068 v.val = sk->sk_reuse;
1069 break;
1070
055dc21a
TH		 1071 case SO_REUSEPORT:
1072 v.val = sk->sk_reuseport;
1073 break;
1074
e71a4783 1075 case SO_KEEPALIVE:
1b23a5df 1076 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1077 break;
1078
1079 case SO_TYPE:
1080 v.val = sk->sk_type;
1081 break;
1082
49c794e9
JE		 1083 case SO_PROTOCOL:
1084 v.val = sk->sk_protocol;
1085 break;
1086
0d6038ee
JE		 1087 case SO_DOMAIN:
1088 v.val = sk->sk_family;
1089 break;
1090
e71a4783
SH		 1091 case SO_ERROR:
1092 v.val = -sock_error(sk);
2a91525c 1093 if (v.val == 0)
e71a4783
SH		 1094 v.val = xchg(&sk->sk_err_soft, 0);
1095 break;
1096
1097 case SO_OOBINLINE:
1b23a5df 1098 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1099 break;
1100
1101 case SO_NO_CHECK:
28448b80 1102 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1103 break;
1104
1105 case SO_PRIORITY:
1106 v.val = sk->sk_priority;
1107 break;
1108
1109 case SO_LINGER:
1110 lv = sizeof(v.ling);
1b23a5df 1111 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1112 v.ling.l_linger = sk->sk_lingertime / HZ;
1113 break;
1114
1115 case SO_BSDCOMPAT:
1116 sock_warn_obsolete_bsdism("getsockopt");
1117 break;
1118
1119 case SO_TIMESTAMP:
92f37fd2
ED		 1120 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1121 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1122 break;
1123
1124 case SO_TIMESTAMPNS:
1125 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1126 break;
1127
20d49473 1128 case SO_TIMESTAMPING:
b9f40e21 1129 v.val = sk->sk_tsflags;
20d49473
PO		 1130 break;
1131
e71a4783 1132 case SO_RCVTIMEO:
2a91525c 1133 lv = sizeof(struct timeval);
e71a4783
SH		 1134 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1135 v.tm.tv_sec = 0;
1136 v.tm.tv_usec = 0;
1137 } else {
1138 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1139 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1140 }
1141 break;
1142
1143 case SO_SNDTIMEO:
2a91525c 1144 lv = sizeof(struct timeval);
e71a4783
SH		 1145 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1146 v.tm.tv_sec = 0;
1147 v.tm.tv_usec = 0;
1148 } else {
1149 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1150 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1151 }
1152 break;
1da177e4 1153
e71a4783
SH		 1154 case SO_RCVLOWAT:
1155 v.val = sk->sk_rcvlowat;
1156 break;
1da177e4 1157
e71a4783 1158 case SO_SNDLOWAT:
2a91525c 1159 v.val = 1;
e71a4783 1160 break;
1da177e4 1161
e71a4783 1162 case SO_PASSCRED:
82981930 1163 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1164 break;
1da177e4 1165
e71a4783 1166 case SO_PEERCRED:
109f6e39
EB		 1167 {
1168 struct ucred peercred;
1169 if (len > sizeof(peercred))
1170 len = sizeof(peercred);
1171 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1172 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1173 return -EFAULT;
1174 goto lenout;
109f6e39 1175 }
1da177e4 1176
e71a4783
SH		 1177 case SO_PEERNAME:
1178 {
1179 char address[128];
1180
1181 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1182 return -ENOTCONN;
1183 if (lv < len)
1184 return -EINVAL;
1185 if (copy_to_user(optval, address, len))
1186 return -EFAULT;
1187 goto lenout;
1188 }
1da177e4 1189
e71a4783
SH		 1190 /* Dubious BSD thing... Probably nobody even uses it, but
1191 * the UNIX standard wants it for whatever reason... -DaveM
1192 */
1193 case SO_ACCEPTCONN:
1194 v.val = sk->sk_state == TCP_LISTEN;
1195 break;
1da177e4 1196
e71a4783 1197 case SO_PASSSEC:
82981930 1198 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1199 break;
877ce7c1 1200
e71a4783
SH		 1201 case SO_PEERSEC:
1202 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1203
4a19ec58
LAT		 1204 case SO_MARK:
1205 v.val = sk->sk_mark;
1206 break;
1207
3b885787 1208 case SO_RXQ_OVFL:
1b23a5df 1209 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1210 break;
1211
6e3e939f 1212 case SO_WIFI_STATUS:
1b23a5df 1213 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1214 break;
1215
ef64a54f
PE		 1216 case SO_PEEK_OFF:
1217 if (!sock->ops->set_peek_off)
1218 return -EOPNOTSUPP;
1219
1220 v.val = sk->sk_peek_off;
1221 break;
bc2f7996 1222 case SO_NOFCS:
1b23a5df 1223 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1224 break;
c91f6df2 1225
f7b86bfe 1226 case SO_BINDTODEVICE:
c91f6df2
BH		 1227 return sock_getbindtodevice(sk, optval, optlen, len);
1228
a8fc9277
PE		 1229 case SO_GET_FILTER:
1230 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1231 if (len < 0)
1232 return len;
1233
1234 goto lenout;
c91f6df2 1235
d59577b6
VB		 1236 case SO_LOCK_FILTER:
1237 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1238 break;
1239
ea02f941
MS		 1240 case SO_BPF_EXTENSIONS:
1241 v.val = bpf_tell_extensions();
1242 break;
1243
7d4c04fc
KJ		 1244 case SO_SELECT_ERR_QUEUE:
1245 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1246 break;
1247
e0d1095a 1248#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1249 case SO_BUSY_POLL:
dafcc438
ET		 1250 v.val = sk->sk_ll_usec;
1251 break;
1252#endif
1253
62748f32
ED		 1254 case SO_MAX_PACING_RATE:
1255 v.val = sk->sk_max_pacing_rate;
1256 break;
1257
2c8c56e1
ED		 1258 case SO_INCOMING_CPU:
1259 v.val = sk->sk_incoming_cpu;
1260 break;
1261
e71a4783 1262 default:
443b5991
YH		 1263 /* We implement the SO_SNDLOWAT etc to not be settable
1264 * (1003.1g 7).
1265 */
e71a4783 1266 return -ENOPROTOOPT;
1da177e4 1267 }
e71a4783 1268
1da177e4
LT		 1269 if (len > lv)
1270 len = lv;
1271 if (copy_to_user(optval, &v, len))
1272 return -EFAULT;
1273lenout:
4ec93edb
YH		 1274 if (put_user(len, optlen))
1275 return -EFAULT;
1276 return 0;
1da177e4
LT		 1277}
1278
a5b5bb9a
IM		 1279/*
1280 * Initialize an sk_lock.
1281 *
1282 * (We also register the sk_lock with the lock validator.)
1283 */
b6f99a21 1284static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1285{
ed07536e
PZ		 1286 sock_lock_init_class_and_name(sk,
1287 af_family_slock_key_strings[sk->sk_family],
1288 af_family_slock_keys + sk->sk_family,
1289 af_family_key_strings[sk->sk_family],
1290 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1291}
1292
4dc6dc71
ED		 1293/*
1294 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1295 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1296 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1297 */
f1a6c4da
PE		 1298static void sock_copy(struct sock *nsk, const struct sock *osk)
1299{
1300#ifdef CONFIG_SECURITY_NETWORK
1301 void *sptr = nsk->sk_security;
1302#endif
68835aba
ED		 1303 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1304
1305 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1306 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1307
f1a6c4da
PE		 1308#ifdef CONFIG_SECURITY_NETWORK
1309 nsk->sk_security = sptr;
1310 security_sk_clone(osk, nsk);
1311#endif
1312}
1313
fcbdf09d
OP		 1314void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1315{
1316 unsigned long nulls1, nulls2;
1317
1318 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1319 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1320 if (nulls1 > nulls2)
1321 swap(nulls1, nulls2);
1322
1323 if (nulls1 != 0)
1324 memset((char *)sk, 0, nulls1);
1325 memset((char *)sk + nulls1 + sizeof(void *), 0,
1326 nulls2 - nulls1 - sizeof(void *));
1327 memset((char *)sk + nulls2 + sizeof(void *), 0,
1328 size - nulls2 - sizeof(void *));
1329}
1330EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1331
2e4afe7b
PE		 1332static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1333 int family)
c308c1b2
PE		 1334{
1335 struct sock *sk;
1336 struct kmem_cache *slab;
1337
1338 slab = prot->slab;
e912b114
ED		 1339 if (slab != NULL) {
1340 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1341 if (!sk)
1342 return sk;
1343 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1344 if (prot->clear_sk)
1345 prot->clear_sk(sk, prot->obj_size);
1346 else
1347 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1348 }
fcbdf09d 1349 } else
c308c1b2
PE		 1350 sk = kmalloc(prot->obj_size, priority);
1351
2e4afe7b 1352 if (sk != NULL) {
a98b65a3
VN		 1353 kmemcheck_annotate_bitfield(sk, flags);
1354
2e4afe7b
PE		 1355 if (security_sk_alloc(sk, family, priority))
1356 goto out_free;
1357
1358 if (!try_module_get(prot->owner))
1359 goto out_free_sec;
e022f0b4 1360 sk_tx_queue_clear(sk);
bd1060a1 1361 cgroup_sk_alloc(&sk->sk_cgrp_data);
2e4afe7b
PE		 1362 }
1363
c308c1b2 1364 return sk;
2e4afe7b
PE		 1365
1366out_free_sec:
1367 security_sk_free(sk);
1368out_free:
1369 if (slab != NULL)
1370 kmem_cache_free(slab, sk);
1371 else
1372 kfree(sk);
1373 return NULL;
c308c1b2
PE		 1374}
1375
1376static void sk_prot_free(struct proto *prot, struct sock *sk)
1377{
1378 struct kmem_cache *slab;
2e4afe7b 1379 struct module *owner;
c308c1b2 1380
2e4afe7b 1381 owner = prot->owner;
c308c1b2 1382 slab = prot->slab;
2e4afe7b 1383
bd1060a1 1384 cgroup_sk_free(&sk->sk_cgrp_data);
2e4afe7b 1385 security_sk_free(sk);
c308c1b2
PE		 1386 if (slab != NULL)
1387 kmem_cache_free(slab, sk);
1388 else
1389 kfree(sk);
2e4afe7b 1390 module_put(owner);
c308c1b2
PE		 1391}
1392
1da177e4
LT		 1393/**
1394 * sk_alloc - All socket objects are allocated here
c4ea43c5 1395 * @net: the applicable net namespace
4dc3b16b
PP		 1396 * @family: protocol family
1397 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1398 * @prot: struct proto associated with this new sock instance
11aa9c28 1399 * @kern: is this to be a kernel socket?
1da177e4 1400 */
1b8d7ae4 1401struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1402 struct proto *prot, int kern)
1da177e4 1403{
c308c1b2 1404 struct sock *sk;
1da177e4 1405
154adbc8 1406 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1407 if (sk) {
154adbc8
PE		 1408 sk->sk_family = family;
1409 /*
1410 * See comment in struct sock definition to understand
1411 * why we need sk_prot_creator -acme
1412 */
1413 sk->sk_prot = sk->sk_prot_creator = prot;
1414 sock_lock_init(sk);
26abe143
EB		 1415 sk->sk_net_refcnt = kern ? 0 : 1;
1416 if (likely(sk->sk_net_refcnt))
1417 get_net(net);
1418 sock_net_set(sk, net);
d66ee058 1419 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1420
2a56a1fe
TH		 1421 sock_update_classid(&sk->sk_cgrp_data);
1422 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1423 }
a79af59e 1424
2e4afe7b 1425 return sk;
1da177e4 1426}
2a91525c 1427EXPORT_SYMBOL(sk_alloc);
1da177e4 1428
a4298e45
ED		 1429/* Sockets having SOCK_RCU_FREE will call this function after one RCU
1430 * grace period. This is the case for UDP sockets and TCP listeners.
1431 */
1432static void __sk_destruct(struct rcu_head *head)
1da177e4 1433{
a4298e45 1434 struct sock *sk = container_of(head, struct sock, sk_rcu);
1da177e4 1435 struct sk_filter *filter;
1da177e4
LT		 1436
1437 if (sk->sk_destruct)
1438 sk->sk_destruct(sk);
1439
a898def2
PM		 1440 filter = rcu_dereference_check(sk->sk_filter,
1441 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1442 if (filter) {
309dd5fc 1443 sk_filter_uncharge(sk, filter);
a9b3cd7f 1444 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4 1445 }
538950a1
CG		 1446 if (rcu_access_pointer(sk->sk_reuseport_cb))
1447 reuseport_detach_sock(sk);
1da177e4 1448
08e29af3 1449 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1450
1451 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1452 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1453 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1454
109f6e39
EB		 1455 if (sk->sk_peer_cred)
1456 put_cred(sk->sk_peer_cred);
1457 put_pid(sk->sk_peer_pid);
26abe143
EB		 1458 if (likely(sk->sk_net_refcnt))
1459 put_net(sock_net(sk));
c308c1b2 1460 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1461}
2b85a34e 1462
a4298e45
ED		 1463void sk_destruct(struct sock *sk)
1464{
1465 if (sock_flag(sk, SOCK_RCU_FREE))
1466 call_rcu(&sk->sk_rcu, __sk_destruct);
1467 else
1468 __sk_destruct(&sk->sk_rcu);
1469}
1470
eb4cb008
CG		 1471static void __sk_free(struct sock *sk)
1472{
b922622e 1473 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1474 sock_diag_broadcast_destroy(sk);
1475 else
1476 sk_destruct(sk);
1477}
1478
2b85a34e
ED		 1479void sk_free(struct sock *sk)
1480{
1481 /*
25985edc 1482 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1483 * some packets are still in some tx queue.
1484 * If not null, sock_wfree() will call __sk_free(sk) later
1485 */
1486 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1487 __sk_free(sk);
1488}
2a91525c 1489EXPORT_SYMBOL(sk_free);
1da177e4 1490
e56c57d0
ED		 1491/**
1492 * sk_clone_lock - clone a socket, and lock its clone
1493 * @sk: the socket to clone
1494 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1495 *
1496 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1497 */
1498struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1499{
8fd1d178 1500 struct sock *newsk;
278571ba 1501 bool is_charged = true;
87d11ceb 1502
8fd1d178 1503 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1504 if (newsk != NULL) {
1505 struct sk_filter *filter;
1506
892c141e 1507 sock_copy(newsk, sk);
87d11ceb
ACM		 1508
1509 /* SANITY */
8a681736
SV		 1510 if (likely(newsk->sk_net_refcnt))
1511 get_net(sock_net(newsk));
87d11ceb
ACM		 1512 sk_node_init(&newsk->sk_node);
1513 sock_lock_init(newsk);
1514 bh_lock_sock(newsk);
fa438ccf 1515 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1516 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1517
1518 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1519 /*
1520 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1521 */
1522 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1523 atomic_set(&newsk->sk_omem_alloc, 0);
1524 skb_queue_head_init(&newsk->sk_receive_queue);
1525 skb_queue_head_init(&newsk->sk_write_queue);
1526
87d11ceb 1527 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1528 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1529 af_callback_keys + newsk->sk_family,
1530 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1531
1532 newsk->sk_dst_cache = NULL;
1533 newsk->sk_wmem_queued = 0;
1534 newsk->sk_forward_alloc = 0;
9caad864 1535 atomic_set(&newsk->sk_drops, 0);
87d11ceb 1536 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1537 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1538
1539 sock_reset_flag(newsk, SOCK_DONE);
1540 skb_queue_head_init(&newsk->sk_error_queue);
1541
0d7da9dd 1542 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1543 if (filter != NULL)
278571ba
AS		 1544 /* though it's an empty new sock, the charging may fail
1545 * if sysctl_optmem_max was changed between creation of
1546 * original socket and cloning
1547 */
1548 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1549
d188ba86 1550 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
87d11ceb
ACM		 1551 /* It is still raw copy of parent, so invalidate
1552 * destructor and make plain sk_free() */
1553 newsk->sk_destruct = NULL;
b0691c8e 1554 bh_unlock_sock(newsk);
87d11ceb
ACM		 1555 sk_free(newsk);
1556 newsk = NULL;
1557 goto out;
1558 }
fa463497 1559 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
87d11ceb
ACM		 1560
1561 newsk->sk_err = 0;
1562 newsk->sk_priority = 0;
2c8c56e1 1563 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1564 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1565 /*
1566 * Before updating sk_refcnt, we must commit prior changes to memory
1567 * (Documentation/RCU/rculist_nulls.txt for details)
1568 */
1569 smp_wmb();
87d11ceb
ACM		 1570 atomic_set(&newsk->sk_refcnt, 2);
1571
1572 /*
1573 * Increment the counter in the same struct proto as the master
1574 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1575 * is the same as sk->sk_prot->socks, as this field was copied
1576 * with memcpy).
1577 *
1578 * This _changes_ the previous behaviour, where
1579 * tcp_create_openreq_child always was incrementing the
1580 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1581 * to be taken into account in all callers. -acme
1582 */
1583 sk_refcnt_debug_inc(newsk);
972692e0 1584 sk_set_socket(newsk, NULL);
43815482 1585 newsk->sk_wq = NULL;
87d11ceb 1586
baac50bb 1587 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
3d596f7b 1588 sock_update_memcg(newsk);
f3f511e1 1589
87d11ceb 1590 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1591 sk_sockets_allocated_inc(newsk);
704da560 1592
080a270f
HFS		 1593 if (sock_needs_netstamp(sk) &&
1594 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1595 net_enable_timestamp();
87d11ceb
ACM		 1596 }
1597out:
1598 return newsk;
1599}
e56c57d0 1600EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1601
9958089a
AK		 1602void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1603{
d6a4e26a
ED		 1604 u32 max_segs = 1;
1605
6bd4f355 1606 sk_dst_set(sk, dst);
9958089a
AK		 1607 sk->sk_route_caps = dst->dev->features;
1608 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1609 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1610 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1611 if (sk_can_gso(sk)) {
82cc1a7a 1612 if (dst->header_len) {
9958089a 1613 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1614 } else {
9958089a 1615 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1616 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1617 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1618 }
9958089a 1619 }
d6a4e26a 1620 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1621}
1622EXPORT_SYMBOL_GPL(sk_setup_caps);
1623
1da177e4
LT		 1624/*
1625 * Simple resource managers for sockets.
1626 */
1627
1628
4ec93edb
YH		 1629/*
1630 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1631 */
1632void sock_wfree(struct sk_buff *skb)
1633{
1634 struct sock *sk = skb->sk;
d99927f4 1635 unsigned int len = skb->truesize;
1da177e4 1636
d99927f4
ED		 1637 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1638 /*
1639 * Keep a reference on sk_wmem_alloc, this will be released
1640 * after sk_write_space() call
1641 */
1642 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1643 sk->sk_write_space(sk);
d99927f4
ED		 1644 len = 1;
1645 }
2b85a34e 1646 /*
d99927f4
ED		 1647 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1648 * could not do because of in-flight packets
2b85a34e 1649 */
d99927f4 1650 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1651 __sk_free(sk);
1da177e4 1652}
2a91525c 1653EXPORT_SYMBOL(sock_wfree);
1da177e4 1654
9e17f8a4
ED		 1655void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1656{
1657 skb_orphan(skb);
1658 skb->sk = sk;
1659#ifdef CONFIG_INET
1660 if (unlikely(!sk_fullsock(sk))) {
1661 skb->destructor = sock_edemux;
1662 sock_hold(sk);
1663 return;
1664 }
1665#endif
1666 skb->destructor = sock_wfree;
1667 skb_set_hash_from_sk(skb, sk);
1668 /*
1669 * We used to take a refcount on sk, but following operation
1670 * is enough to guarantee sk_free() wont free this sock until
1671 * all in-flight packets are completed
1672 */
1673 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1674}
1675EXPORT_SYMBOL(skb_set_owner_w);
1676
f2f872f9
ED		 1677void skb_orphan_partial(struct sk_buff *skb)
1678{
1679 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1680 * so we do not completely orphan skb, but transfert all
1681 * accounted bytes but one, to avoid unexpected reorders.
1682 */
1683 if (skb->destructor == sock_wfree
1684#ifdef CONFIG_INET
1685 || skb->destructor == tcp_wfree
1686#endif
1687 ) {
1688 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1689 skb->truesize = 1;
1690 } else {
1691 skb_orphan(skb);
1692 }
1693}
1694EXPORT_SYMBOL(skb_orphan_partial);
1695
4ec93edb
YH		 1696/*
1697 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1698 */
1699void sock_rfree(struct sk_buff *skb)
1700{
1701 struct sock *sk = skb->sk;
d361fd59 1702 unsigned int len = skb->truesize;
1da177e4 1703
d361fd59
ED		 1704 atomic_sub(len, &sk->sk_rmem_alloc);
1705 sk_mem_uncharge(sk, len);
1da177e4 1706}
2a91525c 1707EXPORT_SYMBOL(sock_rfree);
1da177e4 1708
7768eed8
OH		 1709/*
1710 * Buffer destructor for skbs that are not used directly in read or write
1711 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1712 */
62bccb8c
AD		 1713void sock_efree(struct sk_buff *skb)
1714{
1715 sock_put(skb->sk);
1716}
1717EXPORT_SYMBOL(sock_efree);
1718
976d0201 1719kuid_t sock_i_uid(struct sock *sk)
1da177e4 1720{
976d0201 1721 kuid_t uid;
1da177e4 1722
f064af1e 1723 read_lock_bh(&sk->sk_callback_lock);
976d0201 1724 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1725 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1726 return uid;
1727}
2a91525c 1728EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1729
1730unsigned long sock_i_ino(struct sock *sk)
1731{
1732 unsigned long ino;
1733
f064af1e 1734 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1735 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1736 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1737 return ino;
1738}
2a91525c 1739EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1740
1741/*
1742 * Allocate a skb from the socket's send buffer.
1743 */
86a76caf 1744struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1745 gfp_t priority)
1da177e4
LT		 1746{
1747 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1748 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1749 if (skb) {
1750 skb_set_owner_w(skb, sk);
1751 return skb;
1752 }
1753 }
1754 return NULL;
1755}
2a91525c 1756EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1757
4ec93edb 1758/*
1da177e4 1759 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1760 */
dd0fc66f 1761void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1762{
95c96174 1763 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1764 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1765 void *mem;
1766 /* First do the add, to avoid the race if kmalloc
4ec93edb 1767 * might sleep.
1da177e4
LT		 1768 */
1769 atomic_add(size, &sk->sk_omem_alloc);
1770 mem = kmalloc(size, priority);
1771 if (mem)
1772 return mem;
1773 atomic_sub(size, &sk->sk_omem_alloc);
1774 }
1775 return NULL;
1776}
2a91525c 1777EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1778
79e88659
DB		 1779/* Free an option memory block. Note, we actually want the inline
1780 * here as this allows gcc to detect the nullify and fold away the
1781 * condition entirely.
1da177e4 1782 */
79e88659
DB		 1783static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1784 const bool nullify)
1da177e4 1785{
e53da5fb
DM		 1786 if (WARN_ON_ONCE(!mem))
1787 return;
79e88659
DB		 1788 if (nullify)
1789 kzfree(mem);
1790 else
1791 kfree(mem);
1da177e4
LT		 1792 atomic_sub(size, &sk->sk_omem_alloc);
1793}
79e88659
DB		 1794
1795void sock_kfree_s(struct sock *sk, void *mem, int size)
1796{
1797 __sock_kfree_s(sk, mem, size, false);
1798}
2a91525c 1799EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1800
79e88659
DB		 1801void sock_kzfree_s(struct sock *sk, void *mem, int size)
1802{
1803 __sock_kfree_s(sk, mem, size, true);
1804}
1805EXPORT_SYMBOL(sock_kzfree_s);
1806
1da177e4
LT		 1807/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1808 I think, these locks should be removed for datagram sockets.
1809 */
2a91525c 1810static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1811{
1812 DEFINE_WAIT(wait);
1813
9cd3e072 1814 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1815 for (;;) {
1816 if (!timeo)
1817 break;
1818 if (signal_pending(current))
1819 break;
1820 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1821 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1822 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1823 break;
1824 if (sk->sk_shutdown & SEND_SHUTDOWN)
1825 break;
1826 if (sk->sk_err)
1827 break;
1828 timeo = schedule_timeout(timeo);
1829 }
aa395145 1830 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1831 return timeo;
1832}
1833
1834
1835/*
1836 * Generic send/receive buffer handlers
1837 */
1838
4cc7f68d
HX		 1839struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1840 unsigned long data_len, int noblock,
28d64271 1841 int *errcode, int max_page_order)
1da177e4 1842{
2e4e4410 1843 struct sk_buff *skb;
1da177e4
LT		 1844 long timeo;
1845 int err;
1846
1da177e4 1847 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1848 for (;;) {
1da177e4
LT		 1849 err = sock_error(sk);
1850 if (err != 0)
1851 goto failure;
1852
1853 err = -EPIPE;
1854 if (sk->sk_shutdown & SEND_SHUTDOWN)
1855 goto failure;
1856
2e4e4410
ED		 1857 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1858 break;
28d64271 1859
9cd3e072 1860 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1861 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1862 err = -EAGAIN;
1863 if (!timeo)
1da177e4 1864 goto failure;
2e4e4410
ED		 1865 if (signal_pending(current))
1866 goto interrupted;
1867 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1868 }
2e4e4410
ED		 1869 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1870 errcode, sk->sk_allocation);
1871 if (skb)
1872 skb_set_owner_w(skb, sk);
1da177e4
LT		 1873 return skb;
1874
1875interrupted:
1876 err = sock_intr_errno(timeo);
1877failure:
1878 *errcode = err;
1879 return NULL;
1880}
4cc7f68d 1881EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1882
4ec93edb 1883struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1884 int noblock, int *errcode)
1885{
28d64271 1886 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1887}
2a91525c 1888EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1889
39771b12
WB		 1890int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1891 struct sockcm_cookie *sockc)
1892{
3dd17e63
SHY		 1893 u32 tsflags;
1894
39771b12
WB		 1895 switch (cmsg->cmsg_type) {
1896 case SO_MARK:
1897 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1898 return -EPERM;
1899 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1900 return -EINVAL;
1901 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1902 break;
3dd17e63
SHY		 1903 case SO_TIMESTAMPING:
1904 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1905 return -EINVAL;
1906
1907 tsflags = *(u32 *)CMSG_DATA(cmsg);
1908 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
1909 return -EINVAL;
1910
1911 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
1912 sockc->tsflags |= tsflags;
1913 break;
39771b12
WB		 1914 default:
1915 return -EINVAL;
1916 }
1917 return 0;
1918}
1919EXPORT_SYMBOL(__sock_cmsg_send);
1920
f28ea365
EJ		 1921int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1922 struct sockcm_cookie *sockc)
1923{
1924 struct cmsghdr *cmsg;
39771b12 1925 int ret;
f28ea365
EJ		 1926
1927 for_each_cmsghdr(cmsg, msg) {
1928 if (!CMSG_OK(msg, cmsg))
1929 return -EINVAL;
1930 if (cmsg->cmsg_level != SOL_SOCKET)
1931 continue;
39771b12
WB		 1932 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
1933 if (ret)
1934 return ret;
f28ea365
EJ		 1935 }
1936 return 0;
1937}
1938EXPORT_SYMBOL(sock_cmsg_send);
1939
5640f768
ED		 1940/* On 32bit arches, an skb frag is limited to 2^15 */
1941#define SKB_FRAG_PAGE_ORDER get_order(32768)
1942
400dfd3a
ED		 1943/**
1944 * skb_page_frag_refill - check that a page_frag contains enough room
1945 * @sz: minimum size of the fragment we want to get
1946 * @pfrag: pointer to page_frag
82d5e2b8 1947 * @gfp: priority for memory allocation
400dfd3a
ED		 1948 *
1949 * Note: While this allocator tries to use high order pages, there is
1950 * no guarantee that allocations succeed. Therefore, @sz MUST be
1951 * less or equal than PAGE_SIZE.
1952 */
d9b2938a 1953bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1954{
5640f768 1955 if (pfrag->page) {
fe896d18 1956 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 1957 pfrag->offset = 0;
1958 return true;
1959 }
400dfd3a 1960 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1961 return true;
1962 put_page(pfrag->page);
1963 }
1964
d9b2938a
ED		 1965 pfrag->offset = 0;
1966 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1967 /* Avoid direct reclaim but allow kswapd to wake */
1968 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1969 __GFP_COMP | __GFP_NOWARN |
1970 __GFP_NORETRY,
d9b2938a 1971 SKB_FRAG_PAGE_ORDER);
5640f768 1972 if (likely(pfrag->page)) {
d9b2938a 1973 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1974 return true;
1975 }
d9b2938a
ED		 1976 }
1977 pfrag->page = alloc_page(gfp);
1978 if (likely(pfrag->page)) {
1979 pfrag->size = PAGE_SIZE;
1980 return true;
1981 }
400dfd3a
ED		 1982 return false;
1983}
1984EXPORT_SYMBOL(skb_page_frag_refill);
1985
1986bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1987{
1988 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1989 return true;
1990
5640f768
ED		 1991 sk_enter_memory_pressure(sk);
1992 sk_stream_moderate_sndbuf(sk);
1993 return false;
1994}
1995EXPORT_SYMBOL(sk_page_frag_refill);
1996
1da177e4 1997static void __lock_sock(struct sock *sk)
f39234d6
NK		 1998 __releases(&sk->sk_lock.slock)
1999 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2000{
2001 DEFINE_WAIT(wait);
2002
e71a4783 2003 for (;;) {
1da177e4
LT		 2004 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2005 TASK_UNINTERRUPTIBLE);
2006 spin_unlock_bh(&sk->sk_lock.slock);
2007 schedule();
2008 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 2009 if (!sock_owned_by_user(sk))
1da177e4
LT		 2010 break;
2011 }
2012 finish_wait(&sk->sk_lock.wq, &wait);
2013}
2014
2015static void __release_sock(struct sock *sk)
f39234d6
NK		 2016 __releases(&sk->sk_lock.slock)
2017 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2018{
2019 struct sk_buff *skb = sk->sk_backlog.head;
2020
2021 do {
2022 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2023 bh_unlock_sock(sk);
2024
2025 do {
2026 struct sk_buff *next = skb->next;
2027
e4cbb02a 2028 prefetch(next);
7fee226a 2029 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 2030 skb->next = NULL;
c57943a1 2031 sk_backlog_rcv(sk, skb);
1da177e4
LT		 2032
2033 /*
2034 * We are in process context here with softirqs
2035 * disabled, use cond_resched_softirq() to preempt.
2036 * This is safe to do because we've taken the backlog
2037 * queue private:
2038 */
2039 cond_resched_softirq();
2040
2041 skb = next;
2042 } while (skb != NULL);
2043
2044 bh_lock_sock(sk);
e71a4783 2045 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 2046
2047 /*
2048 * Doing the zeroing here guarantee we can not loop forever
2049 * while a wild producer attempts to flood us.
2050 */
2051 sk->sk_backlog.len = 0;
1da177e4
LT		 2052}
2053
2054/**
2055 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2056 * @sk: sock to wait on
2057 * @timeo: for how long
dfbafc99 2058 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2059 *
2060 * Now socket state including sk->sk_err is changed only under lock,
2061 * hence we may omit checks after joining wait queue.
2062 * We check receive queue before schedule() only as optimization;
2063 * it is very likely that release_sock() added new data.
2064 */
dfbafc99 2065int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2066{
2067 int rc;
2068 DEFINE_WAIT(wait);
2069
aa395145 2070 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9cd3e072 2071 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
dfbafc99 2072 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
9cd3e072 2073 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
aa395145 2074 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2075 return rc;
2076}
1da177e4
LT		 2077EXPORT_SYMBOL(sk_wait_data);
2078
3ab224be
HA		 2079/**
2080 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2081 * @sk: socket
2082 * @size: memory size to allocate
2083 * @kind: allocation type
2084 *
2085 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2086 * rmem allocation. This function assumes that protocols which have
2087 * memory_pressure use sk_wmem_queued as write buffer accounting.
2088 */
2089int __sk_mem_schedule(struct sock *sk, int size, int kind)
2090{
2091 struct proto *prot = sk->sk_prot;
2092 int amt = sk_mem_pages(size);
8d987e5c 2093 long allocated;
3ab224be
HA		 2094
2095 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2096
e805605c
JW		 2097 allocated = sk_memory_allocated_add(sk, amt);
2098
baac50bb
JW		 2099 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2100 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
e805605c 2101 goto suppress_allocation;
3ab224be
HA		 2102
2103 /* Under limit. */
e805605c 2104 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2105 sk_leave_memory_pressure(sk);
3ab224be
HA		 2106 return 1;
2107 }
2108
e805605c
JW		 2109 /* Under pressure. */
2110 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2111 sk_enter_memory_pressure(sk);
3ab224be 2112
e805605c
JW		 2113 /* Over hard limit. */
2114 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2115 goto suppress_allocation;
2116
2117 /* guarantee minimum buffer size under pressure */
2118 if (kind == SK_MEM_RECV) {
2119 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2120 return 1;
180d8cd9 2121
3ab224be
HA		 2122 } else { /* SK_MEM_SEND */
2123 if (sk->sk_type == SOCK_STREAM) {
2124 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2125 return 1;
2126 } else if (atomic_read(&sk->sk_wmem_alloc) <
2127 prot->sysctl_wmem[0])
2128 return 1;
2129 }
2130
180d8cd9 2131 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2132 int alloc;
2133
180d8cd9 2134 if (!sk_under_memory_pressure(sk))
1748376b 2135 return 1;
180d8cd9
GC		 2136 alloc = sk_sockets_allocated_read_positive(sk);
2137 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2138 sk_mem_pages(sk->sk_wmem_queued +
2139 atomic_read(&sk->sk_rmem_alloc) +
2140 sk->sk_forward_alloc))
2141 return 1;
2142 }
2143
2144suppress_allocation:
2145
2146 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2147 sk_stream_moderate_sndbuf(sk);
2148
2149 /* Fail only if socket is _under_ its sndbuf.
2150 * In this case we cannot block, so that we have to fail.
2151 */
2152 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2153 return 1;
2154 }
2155
3847ce32
SM		 2156 trace_sock_exceed_buf_limit(sk, prot, allocated);
2157
3ab224be
HA		 2158 /* Alas. Undo changes. */
2159 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2160
0e90b31f 2161 sk_memory_allocated_sub(sk, amt);
180d8cd9 2162
baac50bb
JW		 2163 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2164 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2165
3ab224be
HA		 2166 return 0;
2167}
3ab224be
HA		 2168EXPORT_SYMBOL(__sk_mem_schedule);
2169
2170/**
69dba9bb 2171 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2172 * @sk: socket
1a24e04e 2173 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2174 */
1a24e04e 2175void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2176{
1a24e04e
ED		 2177 amount >>= SK_MEM_QUANTUM_SHIFT;
2178 sk_memory_allocated_sub(sk, amount);
2179 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2180
baac50bb
JW		 2181 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2182 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2183
180d8cd9
GC		 2184 if (sk_under_memory_pressure(sk) &&
2185 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2186 sk_leave_memory_pressure(sk);
3ab224be 2187}
3ab224be
HA		 2188EXPORT_SYMBOL(__sk_mem_reclaim);
2189
2190
1da177e4
LT		 2191/*
2192 * Set of default routines for initialising struct proto_ops when
2193 * the protocol does not support a particular function. In certain
2194 * cases where it makes no sense for a protocol to have a "do nothing"
2195 * function, some default processing is provided.
2196 */
2197
2198int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2199{
2200 return -EOPNOTSUPP;
2201}
2a91525c 2202EXPORT_SYMBOL(sock_no_bind);
1da177e4 2203
4ec93edb 2204int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2205 int len, int flags)
2206{
2207 return -EOPNOTSUPP;
2208}
2a91525c 2209EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2210
2211int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2212{
2213 return -EOPNOTSUPP;
2214}
2a91525c 2215EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2216
2217int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2218{
2219 return -EOPNOTSUPP;
2220}
2a91525c 2221EXPORT_SYMBOL(sock_no_accept);
1da177e4 2222
4ec93edb 2223int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2224 int *len, int peer)
2225{
2226 return -EOPNOTSUPP;
2227}
2a91525c 2228EXPORT_SYMBOL(sock_no_getname);
1da177e4 2229
2a91525c 2230unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2231{
2232 return 0;
2233}
2a91525c 2234EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2235
2236int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2237{
2238 return -EOPNOTSUPP;
2239}
2a91525c 2240EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2241
2242int sock_no_listen(struct socket *sock, int backlog)
2243{
2244 return -EOPNOTSUPP;
2245}
2a91525c 2246EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2247
2248int sock_no_shutdown(struct socket *sock, int how)
2249{
2250 return -EOPNOTSUPP;
2251}
2a91525c 2252EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2253
2254int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2255 char __user *optval, unsigned int optlen)
1da177e4
LT		 2256{
2257 return -EOPNOTSUPP;
2258}
2a91525c 2259EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2260
2261int sock_no_getsockopt(struct socket *sock, int level, int optname,
2262 char __user *optval, int __user *optlen)
2263{
2264 return -EOPNOTSUPP;
2265}
2a91525c 2266EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2267
1b784140 2268int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2269{
2270 return -EOPNOTSUPP;
2271}
2a91525c 2272EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2273
1b784140
YX		 2274int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2275 int flags)
1da177e4
LT		 2276{
2277 return -EOPNOTSUPP;
2278}
2a91525c 2279EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2280
2281int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2282{
2283 /* Mirror missing mmap method error code */
2284 return -ENODEV;
2285}
2a91525c 2286EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2287
2288ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2289{
2290 ssize_t res;
2291 struct msghdr msg = {.msg_flags = flags};
2292 struct kvec iov;
2293 char *kaddr = kmap(page);
2294 iov.iov_base = kaddr + offset;
2295 iov.iov_len = size;
2296 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2297 kunmap(page);
2298 return res;
2299}
2a91525c 2300EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2301
2302/*
2303 * Default Socket Callbacks
2304 */
2305
2306static void sock_def_wakeup(struct sock *sk)
2307{
43815482
ED		 2308 struct socket_wq *wq;
2309
2310 rcu_read_lock();
2311 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2312 if (skwq_has_sleeper(wq))
43815482
ED		 2313 wake_up_interruptible_all(&wq->wait);
2314 rcu_read_unlock();
1da177e4
LT		 2315}
2316
2317static void sock_def_error_report(struct sock *sk)
2318{
43815482
ED		 2319 struct socket_wq *wq;
2320
2321 rcu_read_lock();
2322 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2323 if (skwq_has_sleeper(wq))
43815482 2324 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2325 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2326 rcu_read_unlock();
1da177e4
LT		 2327}
2328
676d2369 2329static void sock_def_readable(struct sock *sk)
1da177e4 2330{
43815482
ED		 2331 struct socket_wq *wq;
2332
2333 rcu_read_lock();
2334 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2335 if (skwq_has_sleeper(wq))
2c6607c6 2336 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2337 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2338 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2339 rcu_read_unlock();
1da177e4
LT		 2340}
2341
2342static void sock_def_write_space(struct sock *sk)
2343{
43815482
ED		 2344 struct socket_wq *wq;
2345
2346 rcu_read_lock();
1da177e4
LT		 2347
2348 /* Do not wake up a writer until he can make "significant"
2349 * progress. --DaveM
2350 */
e71a4783 2351 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2352 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2353 if (skwq_has_sleeper(wq))
43815482 2354 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2355 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2356
2357 /* Should agree with poll, otherwise some programs break */
2358 if (sock_writeable(sk))
8d8ad9d7 2359 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2360 }
2361
43815482 2362 rcu_read_unlock();
1da177e4
LT		 2363}
2364
2365static void sock_def_destruct(struct sock *sk)
2366{
1da177e4
LT		 2367}
2368
2369void sk_send_sigurg(struct sock *sk)
2370{
2371 if (sk->sk_socket && sk->sk_socket->file)
2372 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2373 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2374}
2a91525c 2375EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2376
2377void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2378 unsigned long expires)
2379{
2380 if (!mod_timer(timer, expires))
2381 sock_hold(sk);
2382}
1da177e4
LT		 2383EXPORT_SYMBOL(sk_reset_timer);
2384
2385void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2386{
25cc4ae9 2387 if (del_timer(timer))
1da177e4
LT		 2388 __sock_put(sk);
2389}
1da177e4
LT		 2390EXPORT_SYMBOL(sk_stop_timer);
2391
2392void sock_init_data(struct socket *sock, struct sock *sk)
2393{
2394 skb_queue_head_init(&sk->sk_receive_queue);
2395 skb_queue_head_init(&sk->sk_write_queue);
2396 skb_queue_head_init(&sk->sk_error_queue);
2397
2398 sk->sk_send_head = NULL;
2399
2400 init_timer(&sk->sk_timer);
4ec93edb 2401
1da177e4
LT		 2402 sk->sk_allocation = GFP_KERNEL;
2403 sk->sk_rcvbuf = sysctl_rmem_default;
2404 sk->sk_sndbuf = sysctl_wmem_default;
2405 sk->sk_state = TCP_CLOSE;
972692e0 2406 sk_set_socket(sk, sock);
1da177e4
LT		 2407
2408 sock_set_flag(sk, SOCK_ZAPPED);
2409
e71a4783 2410 if (sock) {
1da177e4 2411 sk->sk_type = sock->type;
43815482 2412 sk->sk_wq = sock->wq;
1da177e4
LT		 2413 sock->sk = sk;
2414 } else
43815482 2415 sk->sk_wq = NULL;
1da177e4 2416
1da177e4 2417 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2418 lockdep_set_class_and_name(&sk->sk_callback_lock,
2419 af_callback_keys + sk->sk_family,
2420 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2421
2422 sk->sk_state_change = sock_def_wakeup;
2423 sk->sk_data_ready = sock_def_readable;
2424 sk->sk_write_space = sock_def_write_space;
2425 sk->sk_error_report = sock_def_error_report;
2426 sk->sk_destruct = sock_def_destruct;
2427
5640f768
ED		 2428 sk->sk_frag.page = NULL;
2429 sk->sk_frag.offset = 0;
ef64a54f 2430 sk->sk_peek_off = -1;
1da177e4 2431
109f6e39
EB		 2432 sk->sk_peer_pid = NULL;
2433 sk->sk_peer_cred = NULL;
1da177e4
LT		 2434 sk->sk_write_pending = 0;
2435 sk->sk_rcvlowat = 1;
2436 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2437 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2438
f37f0afb 2439 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2440
e0d1095a 2441#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2442 sk->sk_napi_id = 0;
64b0dc51 2443 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2444#endif
2445
62748f32 2446 sk->sk_max_pacing_rate = ~0U;
7eec4174 2447 sk->sk_pacing_rate = ~0U;
70da268b 2448 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2449 /*
2450 * Before updating sk_refcnt, we must commit prior changes to memory
2451 * (Documentation/RCU/rculist_nulls.txt for details)
2452 */
2453 smp_wmb();
1da177e4 2454 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2455 atomic_set(&sk->sk_drops, 0);
1da177e4 2456}
2a91525c 2457EXPORT_SYMBOL(sock_init_data);
1da177e4 2458
b5606c2d 2459void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2460{
2461 might_sleep();
a5b5bb9a 2462 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2463 if (sk->sk_lock.owned)
1da177e4 2464 __lock_sock(sk);
d2e9117c 2465 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2466 spin_unlock(&sk->sk_lock.slock);
2467 /*
2468 * The sk_lock has mutex_lock() semantics here:
2469 */
fcc70d5f 2470 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2471 local_bh_enable();
1da177e4 2472}
fcc70d5f 2473EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2474
b5606c2d 2475void release_sock(struct sock *sk)
1da177e4 2476{
a5b5bb9a
IM		 2477 /*
2478 * The sk_lock has mutex_unlock() semantics:
2479 */
2480 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2481
2482 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2483 if (sk->sk_backlog.tail)
2484 __release_sock(sk);
46d3ceab 2485
c3f9b018
ED		 2486 /* Warning : release_cb() might need to release sk ownership,
2487 * ie call sock_release_ownership(sk) before us.
2488 */
46d3ceab
ED		 2489 if (sk->sk_prot->release_cb)
2490 sk->sk_prot->release_cb(sk);
2491
c3f9b018 2492 sock_release_ownership(sk);
a5b5bb9a
IM		 2493 if (waitqueue_active(&sk->sk_lock.wq))
2494 wake_up(&sk->sk_lock.wq);
2495 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2496}
2497EXPORT_SYMBOL(release_sock);
2498
8a74ad60
ED		 2499/**
2500 * lock_sock_fast - fast version of lock_sock
2501 * @sk: socket
2502 *
2503 * This version should be used for very small section, where process wont block
2504 * return false if fast path is taken
2505 * sk_lock.slock locked, owned = 0, BH disabled
2506 * return true if slow path is taken
2507 * sk_lock.slock unlocked, owned = 1, BH enabled
2508 */
2509bool lock_sock_fast(struct sock *sk)
2510{
2511 might_sleep();
2512 spin_lock_bh(&sk->sk_lock.slock);
2513
2514 if (!sk->sk_lock.owned)
2515 /*
2516 * Note : We must disable BH
2517 */
2518 return false;
2519
2520 __lock_sock(sk);
2521 sk->sk_lock.owned = 1;
2522 spin_unlock(&sk->sk_lock.slock);
2523 /*
2524 * The sk_lock has mutex_lock() semantics here:
2525 */
2526 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2527 local_bh_enable();
2528 return true;
2529}
2530EXPORT_SYMBOL(lock_sock_fast);
2531
1da177e4 2532int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2533{
b7aa0bf7 2534 struct timeval tv;
1da177e4 2535 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2536 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2537 tv = ktime_to_timeval(sk->sk_stamp);
2538 if (tv.tv_sec == -1)
1da177e4 2539 return -ENOENT;
b7aa0bf7
ED		 2540 if (tv.tv_sec == 0) {
2541 sk->sk_stamp = ktime_get_real();
2542 tv = ktime_to_timeval(sk->sk_stamp);
2543 }
2544 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2545}
1da177e4
LT		 2546EXPORT_SYMBOL(sock_get_timestamp);
2547
ae40eb1e
ED		 2548int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2549{
2550 struct timespec ts;
2551 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2552 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2553 ts = ktime_to_timespec(sk->sk_stamp);
2554 if (ts.tv_sec == -1)
2555 return -ENOENT;
2556 if (ts.tv_sec == 0) {
2557 sk->sk_stamp = ktime_get_real();
2558 ts = ktime_to_timespec(sk->sk_stamp);
2559 }
2560 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2561}
2562EXPORT_SYMBOL(sock_get_timestampns);
2563
20d49473 2564void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2565{
20d49473 2566 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2567 unsigned long previous_flags = sk->sk_flags;
2568
20d49473
PO		 2569 sock_set_flag(sk, flag);
2570 /*
2571 * we just set one of the two flags which require net
2572 * time stamping, but time stamping might have been on
2573 * already because of the other one
2574 */
080a270f
HFS		 2575 if (sock_needs_netstamp(sk) &&
2576 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2577 net_enable_timestamp();
1da177e4
LT		 2578 }
2579}
1da177e4 2580
cb820f8e
RC		 2581int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2582 int level, int type)
2583{
2584 struct sock_exterr_skb *serr;
364a9e93 2585 struct sk_buff *skb;
cb820f8e
RC		 2586 int copied, err;
2587
2588 err = -EAGAIN;
364a9e93 2589 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2590 if (skb == NULL)
2591 goto out;
2592
2593 copied = skb->len;
2594 if (copied > len) {
2595 msg->msg_flags |= MSG_TRUNC;
2596 copied = len;
2597 }
51f3d02b 2598 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2599 if (err)
2600 goto out_free_skb;
2601
2602 sock_recv_timestamp(msg, sk, skb);
2603
2604 serr = SKB_EXT_ERR(skb);
2605 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2606
2607 msg->msg_flags |= MSG_ERRQUEUE;
2608 err = copied;
2609
cb820f8e
RC		 2610out_free_skb:
2611 kfree_skb(skb);
2612out:
2613 return err;
2614}
2615EXPORT_SYMBOL(sock_recv_errqueue);
2616
1da177e4
LT		 2617/*
2618 * Get a socket option on an socket.
2619 *
2620 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2621 * asynchronous errors should be reported by getsockopt. We assume
2622 * this means if you specify SO_ERROR (otherwise whats the point of it).
2623 */
2624int sock_common_getsockopt(struct socket *sock, int level, int optname,
2625 char __user *optval, int __user *optlen)
2626{
2627 struct sock *sk = sock->sk;
2628
2629 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2630}
1da177e4
LT		 2631EXPORT_SYMBOL(sock_common_getsockopt);
2632
3fdadf7d 2633#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2634int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2635 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2636{
2637 struct sock *sk = sock->sk;
2638
1e51f951 2639 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2640 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2641 optval, optlen);
3fdadf7d
DM		 2642 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2643}
2644EXPORT_SYMBOL(compat_sock_common_getsockopt);
2645#endif
2646
1b784140
YX		 2647int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2648 int flags)
1da177e4
LT		 2649{
2650 struct sock *sk = sock->sk;
2651 int addr_len = 0;
2652 int err;
2653
1b784140 2654 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2655 flags & ~MSG_DONTWAIT, &addr_len);
2656 if (err >= 0)
2657 msg->msg_namelen = addr_len;
2658 return err;
2659}
1da177e4
LT		 2660EXPORT_SYMBOL(sock_common_recvmsg);
2661
2662/*
2663 * Set socket options on an inet socket.
2664 */
2665int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2666 char __user *optval, unsigned int optlen)
1da177e4
LT		 2667{
2668 struct sock *sk = sock->sk;
2669
2670 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2671}
1da177e4
LT		 2672EXPORT_SYMBOL(sock_common_setsockopt);
2673
3fdadf7d 2674#ifdef CONFIG_COMPAT
543d9cfe 2675int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2676 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2677{
2678 struct sock *sk = sock->sk;
2679
543d9cfe
ACM		 2680 if (sk->sk_prot->compat_setsockopt != NULL)
2681 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2682 optval, optlen);
3fdadf7d
DM		 2683 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2684}
2685EXPORT_SYMBOL(compat_sock_common_setsockopt);
2686#endif
2687
1da177e4
LT		 2688void sk_common_release(struct sock *sk)
2689{
2690 if (sk->sk_prot->destroy)
2691 sk->sk_prot->destroy(sk);
2692
2693 /*
2694 * Observation: when sock_common_release is called, processes have
2695 * no access to socket. But net still has.
2696 * Step one, detach it from networking:
2697 *
2698 * A. Remove from hash tables.
2699 */
2700
2701 sk->sk_prot->unhash(sk);
2702
2703 /*
2704 * In this point socket cannot receive new packets, but it is possible
2705 * that some packets are in flight because some CPU runs receiver and
2706 * did hash table lookup before we unhashed socket. They will achieve
2707 * receive queue and will be purged by socket destructor.
2708 *
2709 * Also we still have packets pending on receive queue and probably,
2710 * our own packets waiting in device queues. sock_destroy will drain
2711 * receive queue, but transmitted packets will delay socket destruction
2712 * until the last reference will be released.
2713 */
2714
2715 sock_orphan(sk);
2716
2717 xfrm_sk_free_policy(sk);
2718
e6848976 2719 sk_refcnt_debug_release(sk);
5640f768
ED		 2720
2721 if (sk->sk_frag.page) {
2722 put_page(sk->sk_frag.page);
2723 sk->sk_frag.page = NULL;
2724 }
2725
1da177e4
LT		 2726 sock_put(sk);
2727}
1da177e4
LT		 2728EXPORT_SYMBOL(sk_common_release);
2729
13ff3d6f
PE		 2730#ifdef CONFIG_PROC_FS
2731#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2732struct prot_inuse {
2733 int val[PROTO_INUSE_NR];
2734};
13ff3d6f
PE		 2735
2736static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2737
2738#ifdef CONFIG_NET_NS
2739void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2740{
d6d9ca0f 2741 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2742}
2743EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2744
2745int sock_prot_inuse_get(struct net *net, struct proto *prot)
2746{
2747 int cpu, idx = prot->inuse_idx;
2748 int res = 0;
2749
2750 for_each_possible_cpu(cpu)
2751 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2752
2753 return res >= 0 ? res : 0;
2754}
2755EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2756
2c8c1e72 2757static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2758{
2759 net->core.inuse = alloc_percpu(struct prot_inuse);
2760 return net->core.inuse ? 0 : -ENOMEM;
2761}
2762
2c8c1e72 2763static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2764{
2765 free_percpu(net->core.inuse);
2766}
2767
2768static struct pernet_operations net_inuse_ops = {
2769 .init = sock_inuse_init_net,
2770 .exit = sock_inuse_exit_net,
2771};
2772
2773static __init int net_inuse_init(void)
2774{
2775 if (register_pernet_subsys(&net_inuse_ops))
2776 panic("Cannot initialize net inuse counters");
2777
2778 return 0;
2779}
2780
2781core_initcall(net_inuse_init);
2782#else
1338d466
PE		 2783static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2784
c29a0bc4 2785void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2786{
d6d9ca0f 2787 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2788}
2789EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2790
c29a0bc4 2791int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2792{
2793 int cpu, idx = prot->inuse_idx;
2794 int res = 0;
2795
2796 for_each_possible_cpu(cpu)
2797 res += per_cpu(prot_inuse, cpu).val[idx];
2798
2799 return res >= 0 ? res : 0;
2800}
2801EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2802#endif
13ff3d6f
PE		 2803
2804static void assign_proto_idx(struct proto *prot)
2805{
2806 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2807
2808 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2809 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2810 return;
2811 }
2812
2813 set_bit(prot->inuse_idx, proto_inuse_idx);
2814}
2815
2816static void release_proto_idx(struct proto *prot)
2817{
2818 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2819 clear_bit(prot->inuse_idx, proto_inuse_idx);
2820}
2821#else
2822static inline void assign_proto_idx(struct proto *prot)
2823{
2824}
2825
2826static inline void release_proto_idx(struct proto *prot)
2827{
2828}
2829#endif
2830
0159dfd3
ED		 2831static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2832{
2833 if (!rsk_prot)
2834 return;
2835 kfree(rsk_prot->slab_name);
2836 rsk_prot->slab_name = NULL;
adf78eda
JL		 2837 kmem_cache_destroy(rsk_prot->slab);
2838 rsk_prot->slab = NULL;
0159dfd3
ED		 2839}
2840
2841static int req_prot_init(const struct proto *prot)
2842{
2843 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2844
2845 if (!rsk_prot)
2846 return 0;
2847
2848 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2849 prot->name);
2850 if (!rsk_prot->slab_name)
2851 return -ENOMEM;
2852
2853 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2854 rsk_prot->obj_size, 0,
e96f78ab 2855 prot->slab_flags, NULL);
0159dfd3
ED		 2856
2857 if (!rsk_prot->slab) {
2858 pr_crit("%s: Can't create request sock SLAB cache!\n",
2859 prot->name);
2860 return -ENOMEM;
2861 }
2862 return 0;
2863}
2864
b733c007
PE		 2865int proto_register(struct proto *prot, int alloc_slab)
2866{
1da177e4
LT		 2867 if (alloc_slab) {
2868 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2869 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2870 NULL);
1da177e4
LT		 2871
2872 if (prot->slab == NULL) {
e005d193
JP		 2873 pr_crit("%s: Can't create sock SLAB cache!\n",
2874 prot->name);
60e7663d 2875 goto out;
1da177e4 2876 }
2e6599cb 2877
0159dfd3
ED		 2878 if (req_prot_init(prot))
2879 goto out_free_request_sock_slab;
8feaf0c0 2880
6d6ee43e 2881 if (prot->twsk_prot != NULL) {
faf23422 2882 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2883
7e56b5d6 2884 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2885 goto out_free_request_sock_slab;
2886
6d6ee43e 2887 prot->twsk_prot->twsk_slab =
7e56b5d6 2888 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2889 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2890 0,
52db70dc 2891 prot->slab_flags,
20c2df83 2892 NULL);
6d6ee43e 2893 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2894 goto out_free_timewait_sock_slab_name;
2895 }
1da177e4
LT		 2896 }
2897
36b77a52 2898 mutex_lock(&proto_list_mutex);
1da177e4 2899 list_add(&prot->node, &proto_list);
13ff3d6f 2900 assign_proto_idx(prot);
36b77a52 2901 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2902 return 0;
2903
8feaf0c0 2904out_free_timewait_sock_slab_name:
7e56b5d6 2905 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2906out_free_request_sock_slab:
0159dfd3
ED		 2907 req_prot_cleanup(prot->rsk_prot);
2908
2e6599cb
ACM		 2909 kmem_cache_destroy(prot->slab);
2910 prot->slab = NULL;
b733c007
PE		 2911out:
2912 return -ENOBUFS;
1da177e4 2913}
1da177e4
LT		 2914EXPORT_SYMBOL(proto_register);
2915
2916void proto_unregister(struct proto *prot)
2917{
36b77a52 2918 mutex_lock(&proto_list_mutex);
13ff3d6f 2919 release_proto_idx(prot);
0a3f4358 2920 list_del(&prot->node);
36b77a52 2921 mutex_unlock(&proto_list_mutex);
1da177e4 2922
adf78eda
JL		 2923 kmem_cache_destroy(prot->slab);
2924 prot->slab = NULL;
1da177e4 2925
0159dfd3 2926 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2927
6d6ee43e 2928 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2929 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2930 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2931 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2932 }
1da177e4 2933}
1da177e4
LT		 2934EXPORT_SYMBOL(proto_unregister);
2935
2936#ifdef CONFIG_PROC_FS
1da177e4 2937static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2938 __acquires(proto_list_mutex)
1da177e4 2939{
36b77a52 2940 mutex_lock(&proto_list_mutex);
60f0438a 2941 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2942}
2943
2944static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2945{
60f0438a 2946 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2947}
2948
2949static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2950 __releases(proto_list_mutex)
1da177e4 2951{
36b77a52 2952 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2953}
2954
2955static char proto_method_implemented(const void *method)
2956{
2957 return method == NULL ? 'n' : 'y';
2958}
180d8cd9
GC		 2959static long sock_prot_memory_allocated(struct proto *proto)
2960{
cb75a36c 2961 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2962}
2963
2964static char *sock_prot_memory_pressure(struct proto *proto)
2965{
2966 return proto->memory_pressure != NULL ?
2967 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2968}
1da177e4
LT		 2969
2970static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2971{
180d8cd9 2972
8d987e5c 2973 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2974 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2975 proto->name,
2976 proto->obj_size,
14e943db 2977 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2978 sock_prot_memory_allocated(proto),
2979 sock_prot_memory_pressure(proto),
1da177e4
LT		 2980 proto->max_header,
2981 proto->slab == NULL ? "no" : "yes",
2982 module_name(proto->owner),
2983 proto_method_implemented(proto->close),
2984 proto_method_implemented(proto->connect),
2985 proto_method_implemented(proto->disconnect),
2986 proto_method_implemented(proto->accept),
2987 proto_method_implemented(proto->ioctl),
2988 proto_method_implemented(proto->init),
2989 proto_method_implemented(proto->destroy),
2990 proto_method_implemented(proto->shutdown),
2991 proto_method_implemented(proto->setsockopt),
2992 proto_method_implemented(proto->getsockopt),
2993 proto_method_implemented(proto->sendmsg),
2994 proto_method_implemented(proto->recvmsg),
2995 proto_method_implemented(proto->sendpage),
2996 proto_method_implemented(proto->bind),
2997 proto_method_implemented(proto->backlog_rcv),
2998 proto_method_implemented(proto->hash),
2999 proto_method_implemented(proto->unhash),
3000 proto_method_implemented(proto->get_port),
3001 proto_method_implemented(proto->enter_memory_pressure));
3002}
3003
3004static int proto_seq_show(struct seq_file *seq, void *v)
3005{
60f0438a 3006 if (v == &proto_list)
1da177e4
LT		 3007 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3008 "protocol",
3009 "size",
3010 "sockets",
3011 "memory",
3012 "press",
3013 "maxhdr",
3014 "slab",
3015 "module",
3016 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3017 else
60f0438a 3018 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 3019 return 0;
3020}
3021
f690808e 3022static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 3023 .start = proto_seq_start,
3024 .next = proto_seq_next,
3025 .stop = proto_seq_stop,
3026 .show = proto_seq_show,
3027};
3028
3029static int proto_seq_open(struct inode *inode, struct file *file)
3030{
14e943db
ED		 3031 return seq_open_net(inode, file, &proto_seq_ops,
3032 sizeof(struct seq_net_private));
1da177e4
LT		 3033}
3034
9a32144e 3035static const struct file_operations proto_seq_fops = {
1da177e4
LT		 3036 .owner = THIS_MODULE,
3037 .open = proto_seq_open,
3038 .read = seq_read,
3039 .llseek = seq_lseek,
14e943db
ED		 3040 .release = seq_release_net,
3041};
3042
3043static __net_init int proto_init_net(struct net *net)
3044{
d4beaa66 3045 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3046 return -ENOMEM;
3047
3048 return 0;
3049}
3050
3051static __net_exit void proto_exit_net(struct net *net)
3052{
ece31ffd 3053 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3054}
3055
3056
3057static __net_initdata struct pernet_operations proto_net_ops = {
3058 .init = proto_init_net,
3059 .exit = proto_exit_net,
1da177e4
LT		 3060};
3061
3062static int __init proto_init(void)
3063{
14e943db 3064 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3065}
3066
3067subsys_initcall(proto_init);
3068
3069#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.398987 seconds and 5 git commands to generate.